Sensors are devices which typically convert physical variables which are to be measured to electronic signals which become part of a control system. They consist of two parts: the sensor structure and the package which protects the device from environments which are often hostile.
Size reductions in the sensor structure are nearly always beneficial. They allow cost reductions via batch fabrication techniques which are borrowed from microelectronics, and are now being carried out at sufficiently small sizes that the production of such devices can appropriately be referred to by the term micromechanics. Micromechanical sensors or microsensors can sometimes be combined with cofabricated microelectronics to yield performance improvements, and can result in structures which are identified as smart sensors. Microminiaturization can expand sensor application areas. This is exemplified by physical sensors for biological systems. Blood pressure and blood gas analysis devices must be small to be effective.
The fabrication techniques which are most directly available for microsensor fabrication have their origins in microelectronics. The central difficulty which one experiences with such techniques is based on the fact that sensors are fundamentally three-dimensional structures whereas integrated circuit construction is based on planar processing which is, of course, two-dimensional. Presently, nearly all microsensor construction techniques are adaptations of planar integrated circuit processing with modest three-dimensional extensions. Thus, in wafer to wafer bonded sensors, IC processing is combined with silicon bulk machining and wafer to wafer bonding to achieve microsensor production. In surface micromachining, planar processing and lateral etching are combined to achieve the necessary three-dimensionality. However, three-dimensional fabrication and non-silicon technologies are becoming more important for microsensor development.
Pressure transducers are the most used and therefore the best understood sensors. They fall into two classes: relative or differential devices, and absolute transducers. The absolute sensor has been more actively investigated for microminiaturization via surface micromachining. See, for example, U.S. Pat. Nos. 4,744,863 and 4,853,669 to Guckel, et al., for a discussion of such sensors and sealing techniques to produce absolute sensors.
Production of a sealed cavity or "pill box" sensor requires vacuum sealing and electronic sensing of the device. Pill box behavior and electronic sensing together contribute to device performance. Thus, very small deformations of a mechanically stiff diaphragm or membrane are acceptable if the sensing scheme is sufficiently sensitive. An overpressure stop which is either provided by the device or the package is necessary because increasing pressures cause increasing deflections and will eventually lead to pill box failure.
Polycrystalline silicon (polysilicon) can be used as the deformable membrane, as described in the foregoing patents and in U.S. Pat. No. 4,897,360, the disclosure of which is incorporated herein by reference. A typical polysilicon diaphragm film will be able to support a maximum strain of about 1.5% before it fractures. It will also have a built-in strain field. This strain level must either be controllable and therefore becomes a part of the design process, or the film must have the property that processing techniques exist which cause the built-in strain to disappear. For the case of an absolute pressure transducer as exemplified in U.S. Pat. Nos. 4,744,863 and 4,853,669, if it is assumed that the transducer is square, the maximum pressure range is defined by a diaphragm deflection at the center of the diaphragm which is equal to the cavity depth under the diaphragm. The deflection of the diaphragm under pressure induces diaphragm strain. The strain field maximizes at the clamped edge midway between the corners. This strain value cannot exceed the maximum allowed strain for the diaphragm material, thus defining and limiting the pressure range of the sensor.
A pressure transducer requires a sensing mechanism. Piezoresistive and capacitive techniques form the most direct approaches. Piezoresistive sensing is the most directly implemented technique and profits from the availability of polysilicon which can be used to produce excellent, stable and dielectrically isolated sensing structures. Such structures can be formed by covering the polysilicon diaphragm with a silicon nitride layer, a patterned and doped polysilicon layer, and a protective nitride layer. The issue becomes then one of performance evaluation for particular resistor doping levels and resistor placements.
Polysilicon resistors are quite different from diffused silicon resistors. The piezoresistive effect in these devices is roughly a factor of five smaller than that of a well designed single crystal counterpart. Longitudinal gage factors are typically slightly above 20 and transverse gage factors are near -8. The temperature coefficient of resistance can be positive or negative and can be close to zero. The noise figure for these devices involves only thermal noise, which is normally only found to be true for very good metal film resistors. Polysilicon resistors are dielectrically isolated which allows for higher temperature applications because junction leakage currents are absent.
The placement issue for these devices is again quite different than for single crystal resistor placement. The general rule is simply to locate the resistors in the maximum stress regions on the diaphragm. This would imply longitudinal sections which enter the diaphragm at the support midway between corners. There is, however, a problem. Diaphragm sizes will typically be less than 100 micrometers on a side. The resistors will therefore be quite small, with typical line widths of 4 micrometers. Thus, alignment tolerances as well as line width shifts during polysilicon etching must be considered. A full transducer may use four devices in a bridge configuration, with two resistors which are pressure sensitive and two resistors which are insensitive to pressure because their pill box oxide has not been removed. With this configuration, the half-active full bridge, the maximum output in millivolt per volt of bridge excitation can be calculated at the touch-down pressure.
The difficulties of piezoresistive sensing can be removed by changing the sensing technique or by converting the device from an absolute pressure sensor to a differential transducer. Both approaches are receiving detailed attention. In the first case the piezoresistor can be replaced by a new type of force sensor: the vacuum sealed resonating beam. It is essentially a pill box which contains a doubly clamped free standing beam which can be excited electrostatically. The transduction mechanism is axially applied force to frequency. Its very high sensitivity allows for simpler and more precise measurements. The draw-back is found in the increased complexity of the necessary construction techniques. The device becomes expensive and is of primary use where low pressure precision measurements are required.